Question Advice for deep learning build

DuaneWhitney

Distinguished
Sep 27, 2009
18
2
18,515
I'm building a desktop for training financial/economic deep learning models and need some advice. I currently train on an asus gaming laptop, but it runs very hot and hits memory constraints. This is what I'm considering so far:

300 MSI MAG X870 TOMAHAWK
or
500 MSI MAG X870 CARBON

180 Antec Flux Pro
90 ARCTIC COOLING Liquid Freezer III - 360
383 AMD Ryzen 9 9900X 4.4 GHz 12-Core Processor
330 Corsair Vengeance 96 GB (2 x 48 GB) DDR5-6000 CL30 Memory
130 Corsair RM1000x (2024) 1000 W Fully Modular ATX Power Supply
120 Samsung 990 Pro 2 TB M.2-2280 PCIe 4.0 X4 NVME Solid State Drive
200 Windows Pro
19 Kingpin Cooling KPx Thermal Grease 3g
180 WD Black Sata 8TB Drive


I was planning on waiting on the graphics card till early next year when the new nvidia cards are released. I'll get an inexpensive card for now. With Keras, tensorflow won't train on the GPU anyway in windows. I'll migrate to pytorch (which trains using the GPU) early next year.

I'm leaning toward the Tomahawk, but the extra usb connectivity of the Carbon is beneficial (I tend to run out of ports). I was also considering the Asrock Nova, but seems to be out of stock everywhere.

Any advice is greatly appreciated.

Thanks!

Duane.
 

DuaneWhitney

Distinguished
Sep 27, 2009
18
2
18,515
  • Like
Reactions: Why_Me

DuaneWhitney

Distinguished
Sep 27, 2009
18
2
18,515
Alternative. You can use the integrated graphics that's included with this cpu.

PCPartPicker Part List

CPU: *Intel Core Ultra 7 265K 3.9 GHz 20-Core Processor ($384.99 @ B&H)
Motherboard: *Gigabyte Z890 EAGLE WIFI7 ATX LGA1851 Motherboard ($229.99 @ B&H)
Total: $614.98
Prices include shipping, taxes, and discounts when available
*Lowest price parts chosen from parametric criteria
Generated by PCPartPicker 2024-11-23 23:29 EST-0500
thanks for the suggestion! looks like this processsor is about 20-30% faster for my applications (compared to 9900x).
 
  • Like
Reactions: Why_Me